search

kevaldoshi17 / Prediction-based-Video-Anomaly-Detection-

MIT License
17 stars 10 forks source link

Real Time Code #4

Closed vishalned closed 3 years ago

vishalned commented 3 years ago

Hey, So I have noticed that the notebook that you have created makes use of the pre-generated psnr file. I know that we can create the bounding boxes using YOLO in real-time, but could you maybe release the code or give some suggestions on running the code on real-time, i.e without any pre-generated psnr results. Thanks in advance.

kevaldoshi17 commented 3 years ago

Hi, to run the code in real-time, all you need to do is run the inference of the future frame algorithm for a given batch of input frames (like a sliding window) and use the psnr between the predicted and actual frame with the YOLO bounding boxes.

vishalned commented 3 years ago

got it, thanks.

mausLe commented 3 years ago

Hi, to run the code in real-time, all you need to do is run the inference of the future frame algorithm for a given batch of input frames (like a sliding window) and use the psnr between the predicted and actual frame with the YOLO bounding boxes.

Hi @kevaldoshi17 How could I integrate the inference, YOLO, and the MONAD notebook to run these in real-time?

Zain-Gill123 commented 3 years ago

Hey @mausLe ! Did you find how to run the Inference, YOLO and MONAD notebook in Real-Time? Cheers!

mausLe commented 3 years ago

Hi @Zain-Gill123 I had not found any way to run it in real-time. The average FPS when I ran it on RTX 3090 was about 14 FPS.

Zain-Gill123 commented 3 years ago

Thanks @mausLe ! Thank you for helping. Would mind sharing that notebook with me? You have already shared one with me, is it the same one? Cheers!

mausLe commented 3 years ago

@Zain-Gill123 Yeah it is.